| Summary: | "In many works involving measurement invariance testing, researchers concentrate on one type of grouping only, such as countries, even when the comparisons they make involve multiple types of grouping, such as countries and years. In this article, we propose a procedure allowing to incorporate more than one type of grouping into the invariance testing. For that, we use the example of political participation which is often studied in a comparative perspective where both countries and years are considered. The results show that the comparability of levels of political participation in Europe over the last 20 years is limited. With a simulation study, we show that one remedy for this could be alignment optimization which produces more accurate estimates of means and standard errors. Furthermore, we demonstrate that ignoring the non-invariance can change our substantive conclusions regarding the aggregated trends of participation.
|
|---|